These pages are auto-generated from self-documenting comments embedded in class files.
For more information on breve and steve, refer to the breve homepage.

Abstract : NeuralNetwork

This class is included as part of the file NeuralNetwork.tz.
To use this class in a simulation, include the line "@use NeuralNetwork."

Class description:

A new experimental class for simple feed-forward neural networks.

Class methods:

add-layer with-neuron-count count (int)

Adds a layer to the output end of the neural network, with count neurons. The first layer added serves as the input layer, and then subsequent layers serve as hidden or output layers.

To create a network with one hidden layer, for example, you would first call add-layer for the input layer, then the hidden layer, then the output layer.

get-output at-position position (int)

Returns the value at index position of the output layer. This assumes that run-with has already been called to actually run the neural network.

get-weight on-layer theLayer (int) from-node fn (int) to-node tn (int)

Sets the input to the node number tn on layer theLayer, coming from node number fn on layer number theLayer - 1 to weight. Normally you wouldn't need to use this method, since the method get-output is the preferred way to retrieve network output.

run-with inputs inputList (list)

Sets the inputs of the neural network to the values in inputList, and runs the network. Output values can then be retrieved with get-output.

set-weight to-layer theLayer (int) from-node fn (int) to-node tn (int) to-weight weight (float)

Sets the input to the node number tn on layer theLayer, coming from node number fn on layer number theLayer - 1 to weight.

Documentation created Sun Oct 7 19:13:03 2007